The CFO's tech investment paradox

The paradox playing out in boardrooms

You’re asked for a refreshed cash flow forecast. Again. Faster this time. Every line of “optional spend” is scrutinised.

Then, in the same meeting, you’re asked what the business is doing about AI. The question isn’t “do we invest?”, but “how quickly?”. The board is becoming increasingly nervous about falling behind.

That is the investment paradox.

Most boards want two things at once: certainty and speed.  It’s natural to gravitate toward what feels safest when under pressure, but in practice, that often defaults to business cases predicated on unrealistic headcount savings.

A spreadsheet built on headcount cuts is a comfort blanket, not a long-term tool for defensible decision-making. It narrows the field of vision and pulls technology investment decisions toward short-term efficiencies. You would be saving pennies while missing pounds.

In this insight, Mark O’Sullivan sets out seven critical questions CFOs can bring into t

he boardroom to support investment decisions that will still stand up to scrutiny 12 months from now.

Why AI does not behave like other technology investments

Technology investment decisions follow a familiar logic: define the scope, assess the costs to deliver (one off and business as usual), then model the returns, approve the spend and measure the payback.

AI doesn’t behave that neatly.

• Value shows up in unfamiliar places, it’s often a chain of small improvements that are difficult to quantify
• The capability evolves as you build - what looked cutting edge six months ago is now out of date and you don't have historic evidence of what works
• There aren’t just financial constraints, as with most technology investments there are data quality, risk appetite, operating model readiness, controls, adoption, and internal capacity and capabilities to consider

A no-regrets framework for AI investment decisions

Here are seven questions to help stop your business from drifting into either extreme: reckless spend or decision paralysis.

1) Do we have the operating model for this to work?

In mature organisations, innovation is often stifled by the existing operating model. This isn’t just about internal resistance; for innovation to thrive, it will often rely on different processes, systems and ways of measuring value. It cannot just be bolted on.

No-regrets move: focus on how AI fits in before you decide which tool you buy.

Practical questions:

• Do you have the architecture in place to scale beyond a pilot, across geographies, currencies, and functions?
• Who will own the end-to-end outcome of this investment?
• Are you trying to bolt new tools onto processes that are already broken?

This doesn't mean having all the answers upfront. Agility is key but it does mean having considered a route to scaling solutions.

2) Are you pricing in the cost of inaction?

AI business cases often struggle because leaders focus on what they can quantify today. It’s much harder to measure what you miss by moving too slowly.

Finance leaders recognise this in theory. 84% of CFOs acknowledge that competitors outpacing them with AI could impact their business within the next 12 months. The cost of inaction is rarely quantified.

No-regrets move: treat “lost opportunity” as a real line in the decision, even if it is a range not a solid number.

Practical ways to do that:

• Define the competitive scenarios that would materially change your position (pricing, customer experience, speed to market, cost to serve)
• Fund controlled experimentation so you are building muscle, not just buying tools
• Agree what “being left behind” would look like for your business, in simple terms

Alternatively Speaking

AI strategy: does scrappy speed beat corporate caution?

Listen now

This is risk management, with a long-term lens.

3) Are we making a 24-month decision with a 12-week mindset?

When cash is under pressure, businesses slip into short-horizon decision making. It's easy to see how this happens, but it’s expensive in the long term.

The risk is that short-term, rushed fixes fail to land, and confidence in future investment drops with them.

That makes it harder to secure investment in future cases. Teams disengage from tools they know are temporary and are likely to be replaced months later, creating a cycle of stop-start adoption with little long-term value.

No-regrets move: assess AI investment priorities across three horizons. Be clear on what you are trying to achieve.

A simple structure:

• Now (0–3 months): quick wins that relieve pressure on stretched teams and improve decision speed (for example, automating high-friction reporting steps)
• Next (3–12 months): scalable capability that strengthens forecasting, controls, and insight
• Beyond (12+ months): the “art of the possible” bets that require a fundamental rethink of your operating model

If everything sits in “now,” you get tools that reduce pain but do not build resilience.

4) Is the organisational mindset ready for adoption?

Most businesses now have some form of technology training in place. If that training is mandatory, you will have brilliant completion rates on paper but that might give a distorted picture of your current capabilities.

If people feel the roll out of AI tools is something that is being done to them, adoption always turns into quiet avoidance. If people fear being replaced, they will not experiment. If leaders cannot tolerate ambiguity, every pilot becomes a political battle.

No-regrets move: build psychological safety and clarity, not just capability.

Practical steps:

• Make experimentation feel safe, with clear boundaries and governance
• Proactively reward continuous learning, not just positive outcomes
• Be transparent about how roles will change, and what support people will get

Used effectively, AI will not just make individual tasks more efficient; it will change employees’ identity at work. If you ignore that now, you will inevitably pay for it later. We see most AI programmes do not fail at the point of investment, or because something “goes wrong”; they fail quietly when tools exist, but no one uses them.

5) Are we embedding ethics and controls, or retrofitting them?

AI risk and ethics should not be a footnote. They should be a core part of the decision.

There are governance, security, data, and ethical questions that you won’t be able to outsource to a policy document after procurement has started. So, if the board is asking “what’s the ROI?”, they should also be asking “what’s the governance?”

No-regrets move: articulate effective governance as an enabler, not a brake on pace.

This shouldn't mean endless admin and reporting. It means making sure the risks are considered through every layer of your operating model: from training people and sharpening policies to reviewing security.

A practical approach:

• Make sure there is a clear, shared understanding of your risk appetite
• Identify any hard limits on what your business will not do with AI, based on your business' values and ethics
• Put in place clear accountability for model use, data sources, and sign-off
• Make sure controls will scale, so the business can move faster safely

If you get this right, you won’t slow innovation. Instead, you will help the business to make decisions that will stick and prevent panic later.

6) What is the real driver: a business need, or a fear of missing out?

Every vendor in the market is now “AI-enabled.” The vast majority are selling variations of the same underlying tooling. When decisions are driven by urgency or external pressure, technology investments too often get approved without a clear role.

The no-regrets move: identify the real underlying driver for every business case.

Ask:

• What core business challenge will tools solve? How does that align with our strategy and KPIs?
• Which decisions in our business are high frequency, high friction, or high consequence?
• Which processes are most exposed when volatility hits (cash forecasting, supply chain, pricing, credit control)?

Then work backwards into the tools your business needs. This will need to be continually revisited; AI is not a project you can complete. It is a capability you build continuously.

7) Do we have the capacity and experience to make this investment work?

There is a pattern emerging across many businesses: AI projects are layered on top of ongoing transformation programmes, all landing on teams that are already stretched and absorbing a large volume of change.

While the long-term ambition is to drive productivity through technology, embedding new tools effectively requires focus, capability, and space to adapt upfront. Where multiple initiatives are already in flight, there needs to be challenge around the organisational capacity to take on something new.

The Barometer warns us about this from multiple angles:

• Teams are already under unsustainable stretch (85%)
• Transformation stall risk is high (79%)
• Liquidity pressure is high (81%)

These are not separate problems, they compound.

That is why “no-regrets” AI investment isn’t about doing everything; it often looks like a smaller, targeted set of priorities, executed properly and with a clear roadmap.

Sometimes the most resilient AI investment decision is to pause a lower value project to protect the one that will add the most value.

CFO Forum

A CFO’s technology risk sense check

Sign up